Characteristics of sichrome resistors measured by D. Dowell

The sichrome resistors could be useful for large load resistors for the semiconductive pop-up arrays.

Darren Dowell -- (626)395-6675 (office), -2600 (lab), 796-8806 (FAX)

Last modified Thursday, 12-Dec-2000 10:30 PST
cdd@submm.caltech.edu

Cold noise spectra (SiCr on quartz) -- July 2000

Resistor 3/43 was connected in series with a 30 Mohm nichrome resistor (simulating a bolometer), and resistor 3/44 was connected in series with a 30 Mohm resistor. The outputs of the 2 voltage dividers were routed to the inputs of a differential amplifier.

Cold noise spectrum. Raw data and Postscript figure available here. The rise in noise toward low frequencies for the case of 0 V bias is probably due to amplifier noise. The excess of low frequency noise for f < 1 Hz with the 5 V bias could be due to bias amplitude noise or due to current noise in one of the resistors. The high frequency junk is microphonic response of vibrating cables.

Tests of devices on quartz -- March 2000

Seven devices (6 good) were obtained from M. Jhabvala in Mar. 2000. They are long resistors with patterned SiCr on quartz substrates, glued to a metal package. The box was labeled "HAWC // RL // W2 // Row 2,3 // Run M1 // 3/9/00 MJ".

row/column     R at 300K      comments
----------     ----------     --------
2/6            408.1 Mohm
2/7            402.1
2/8            387.4
2/10           *inf.*         saw cut
3/43           699.1
3/44           691.7
3/45           682.7

Resistance vs. temperature. Raw data and Postscript figure available here.

Noise spectrum with grounded resistor terminal. The noise spectrum of the SiCr on quartz resistor (black curve) is similar to that of a Victoreen MOX resistor (blue curve) of the same value. (The Victoreen spectrum has microphonic spikes which can be ignored.) The spectra roll off as frequency^-1, as expected for a simple RC filter. The corner frequency implies a capacitance from the JFET gate to ground of 23 pF, which could be entirely due to the JFET. For comparison, the unusual frequency dependence of the SiCr on Si resistor is shown (green curve).

There is an indication of some shunt capacitance across the resistor from this test. Otherwise, the output would fall as frequency^-1. However, the shunting effect is not as severe as for the devices with Si substrate (green curve).

The glitches on transitions indicate a small shunt capacitance across the terminals of the load resistor (387 Mohm).

TENTATIVE CONCLUSION: These resistors on quartz are significantly better than the devices on silicon. They are suitable for SHARC2 (and HAWC as far as I understand the requirements). The capacitance from one terminal to the other will hopefully be reduced in the actual array package. I have approximately 20 pF of JFET gate capacitance which limits the knowledge of the capacitance from a resistor terminal to ground.

Tests of 500 Mohm samples at 1.5K -- December 1999

Eight devices (6 good) were obtained from M. Jhabvala in Oct. 1999. The plastic case they were transported in was labeled "HAWC RL 10/99 MDJ NASA/GSFC". The resistors were patterned on silicon dies approximately 1 inch long by MSI. They were glued to a hefty metal package and wire bonded to solder leads. Here are the row/column numbers of the devices:

row/column     location           comments
----------     --------------     --------
3/73           A side, edge
2/78           A side, (next)     lower resistance
2/79           A side, (next)     die damage; tap for ~1/3 of resistor
3/79           A side, middle
               B side, edge       no continuity
3/59           B side, (next)
3/60           B side, (next)
3/64           B side, middle

Resistance vs. temperature for large MSI sichrome resistors. There was no significant variation in resistance with voltage from 0.1 V applied to 30 V applied, except for a drop in resistance at the coldest temperature for V > 10 V. (Probably due to self-heating.) According to MSI, wafer 'A' had 1370 ohms/square, and wafer 'B' had 1200 ohms/square. (Do those 'A' and 'B' designations correspond to the hand-written designations on the package? I assume so.) It appears that the resistance correlates with both wafer designation and row. 'A/3' is highest, 'B/3' is intermediate, and 'A/2' is lowest. The temperature behavior of the resistance is quite similar for all of the devices.

Noise measurements. Four resistors were chosen for noise tests in a bridge configuration (3/73, 3/79, 3/59, 3/60). In the bridge circuit, noise in any one of the four resistors will appear at the output of the amplifier. The resistors and JFETs were located inside an IR Labs HDL-10 dewar and read out with the SHARC2 prototype electronics and a Stanford Research spectrum analyzer. The resistors were cooled to 1.4 K.

Three bias voltages were applied to the bridge: 0 V, 2 V, and 5 V. With a bias of 0 V, only Johnson noise should appear in the spectrum. With a nonzero bias, resistance fluctuations should also be evident in the spectrum. There was no observed voltage dependence of the noise spectrum down to 0.3 Hz, indicating low noise/stable resistors. However, the spectrum begins to roll off at the very low frequency of 3 Hz, presumably due to parasitic capacitance in the vicinity of the resistors. The frequency dependence is f^-1/2 rather than the f^-1 expected for a simple RC lowpass filter. This is potentially caused by the capacitance being distributed.

In order to diagnose the capacitance, the room temperature noise of a single sichrome resistor was compared with that of a Victoreen MOX resistor in a radial lead package. The sichrome resistor had the same f^-1/2 rolloff as at cold temperatures, while the MOX resistor had a simple RC (f^-1) rolloff with a total capacitance of 12 pF. This indicates that the sichrome resistor is responsible for the strange frequency behavior and not the amplifier. A measurement of the short resistor (low value) also displayed the same behavior.

Tests of frequency response (not shown here) indicate large (>10 pF) capacitance from the resistor to ground and across the terminals of the resistor. CONCLUSION: This device is unsuitable for use in SHARC II due to large parasitic capacitance.

Tests of 20 Mohm samples at 1.5K -- May 1999

The samples: 20 Mohm +- 1%, part number MSHR 4SS-20004F-G, date code 9844, lot #A1634C-9

Resistance vs. temperature for two MSI 20 Mohm sichrome resistors in series

Resistance vs. temperature for two MSI 30 Mohm nichrome resistors in series, for comparison

An AC bridge circuit was used to measure the low-frequency Delta(R)/R stability at 1.5K:

The AC waveform is digitized at 1 kHz and demodulated with a DSP (the KAO data system). The noise spectrum for the resulting quasi-DC output, with a +-0.49V bias across the bridge at 50 Hz, is as shown:

The resistance appears to be very stable on time scales up to a few minutes. For comparison, here is the spectrum with the AC component turned off, which should only be sensitive to the noise at 50 Hz:

I don't understand why the noise seems to be ~50% higher than the expected Johnson noise. This could be due to an incorrect conversion of ADU to V, a missing factor of sqrt(2) in the Fourier transform, or noise at frequencies other than 50 Hz entering the system. In any case, the noise for AC bias on and off is the same, which is an indication of resistance stability.

At attempt to confirm the noise with a DC measurement was made. DC bias voltages of 0V and 0.49V were applied to the bridge. The output of the preamplifier was sampled with a spectrum analyzer:

The noise spectra for the non-zero and zero bias overlap, meaning that no current noise is evident. The horizontal line indicates the Johnson noise predicted from the temperature and resistance, which is in very good agreement for frequencies above 5 Hz.

Assuming that the voltage noise due to resistance fluctuations is < 30 nV/sqrt(Hz) at all frequencies, then the limit on delta(r)/r is 9x10^-8/sqrt(Hz). (Calculated with delta(r)/r = sqrt(2) delta(v)/v.)

Tests of 150 Mohm samples at 0.35K -- August 1999

The samples: 150 Mohm +- 1%, part number MSHR 4SS-15005F-E, date code 9924, lot #A1202C-3

Resistance vs. temperature for two MSI 150 Mohm sichrome resistors in series. The two samples had resistances differing by ~7% at the cold temperatures.

The AC bridge circuit was used to measure the low-frequency Delta(R)/R stability at 0.35K. The noise spectrum for the resulting quasi-DC output, with a +-0.42V bias across the bridge at 7 Hz, is as shown:

A scaling factor has been applied to the data to bring the high-frequency white noise to the correct level. The resistance appears to be stable on time scales up to a 10 seconds. For longer time scales, some part of the system has a drift, possibly the bias generator. Here is the output signal prior to Fourier transforming:

DC measurements were also made. DC bias voltages of 0V, 1.5V, and 3.0V were applied to the bridge. The output of the preamplifier was sampled with a spectrum analyzer:

The horizontal line indicates the Johnson noise predicted from the temperature and resistance, which is in good agreement with the noise for 0V bias. For the higher bias voltages, however, there is excess noise. This could be due to bias voltage drifts.

The falling noise at high frequencies is due to capacitive shorting of the signal to the FET gate. The corner frequency is at approximately 70 Hz, which implies a capacitance of 23 pF. A DC impedance of 100 Mohms was assumed.

Go to SHARC II home page...